1Using the glibc microbenchmark suite 2==================================== 3 4The glibc microbenchmark suite automatically generates code for specified 5functions, builds and calls them repeatedly for given inputs to give some 6basic performance properties of the function. 7 8Running the benchmark: 9===================== 10 11The benchmark needs python 2.7 or later in addition to the 12dependencies required to build the GNU C Library. One may run the 13benchmark by invoking make as follows: 14 15 $ make bench 16 17This runs each function for 10 seconds and appends its output to 18benchtests/bench.out. To ensure that the tests are rebuilt, one could run: 19 20 $ make bench-clean 21 22The duration of each test can be configured setting the BENCH_DURATION variable 23in the call to make. One should run `make bench-clean' before changing 24BENCH_DURATION. 25 26 $ make BENCH_DURATION=1 bench 27 28The benchmark suite does function call measurements using architecture-specific 29high precision timing instructions whenever available. When such support is 30not available, it uses clock_gettime (CLOCK_MONOTONIC). One can force the 31benchmark to use clock_gettime by invoking make as follows: 32 33 $ make USE_CLOCK_GETTIME=1 bench 34 35Again, one must run `make bench-clean' before changing the measurement method. 36 37On x86 processors, RDTSCP instruction provides more precise timing data 38than RDTSC instruction. All x86 processors since 2010 support RDTSCP 39instruction. One can force the benchmark to use RDTSCP by invoking make 40as follows: 41 42 $ make USE_RDTSCP=1 bench 43 44One must run `make bench-clean' before changing the measurement method. 45 46Running benchmarks on another target: 47==================================== 48 49If the target where you want to run benchmarks is not capable of building the 50code or you're cross-building, you could build and execute the benchmark in 51separate steps. On the build system run: 52 53 $ make bench-build 54 55and then copy the source and build directories to the target and run the 56benchmarks from the build directory as usual: 57 58 $ make bench 59 60make sure the copy preserves timestamps by using either rsync or scp -p 61otherwise the above command may try to build the benchmark again. Benchmarks 62that require generated code to be executed during the build are skipped when 63cross-building. 64 65Building benchmarks as static executables: 66========================================= 67 68To build benchmarks as static executables, on the build system, run: 69 70 $ make STATIC-BENCHTESTS=yes bench-build 71 72You can copy benchmark executables to another machine and run them 73without copying the source nor build directories. 74 75Running subsets of benchmarks: 76============================== 77 78To run only a subset of benchmarks, one may invoke make as follows 79 80 $ make bench BENCHSET="bench-pthread bench-math malloc-thread" 81 82where BENCHSET may be a space-separated list of the following values: 83 84 bench-math 85 bench-pthread 86 bench-string 87 hash-benchset 88 malloc-thread 89 math-benchset 90 stdio-common-benchset 91 stdlib-benchset 92 string-benchset 93 wcsmbs-benchset 94 95Adding a function to benchtests: 96=============================== 97 98If the name of the function is `foo', then the following procedure should allow 99one to add `foo' to the bench tests: 100 101- Append the function name to the bench variable in the Makefile. 102 103- Make a file called `foo-inputs` to provide the definition and input for the 104 function. The file should have some directives telling the parser script 105 about the function and then one input per line. Directives are lines that 106 have a special meaning for the parser and they begin with two hashes '##'. 107 The following directives are recognized: 108 109 - args: This should be assigned a colon separated list of types of the input 110 arguments. This directive may be skipped if the function does not take any 111 inputs. One may identify output arguments by nesting them in <>. The 112 generator will create variables to get outputs from the calling function. 113 - ret: This should be assigned the type that the function returns. This 114 directive may be skipped if the function does not return a value. 115 - includes: This should be assigned a comma-separated list of headers that 116 need to be included to provide declarations for the function and types it 117 may need (specifically, this includes using "#include <header>"). 118 - include-sources: This should be assigned a comma-separated list of source 119 files that need to be included to provide definitions of global variables 120 and functions (specifically, this includes using "#include "source"). 121 See pthread_once-inputs and pthreads_once-source.c for an example of how 122 to use this to benchmark a function that needs state across several calls. 123 - init: Name of an initializer function to call to initialize the benchtest. 124 - name: See following section for instructions on how to use this directive. 125 126 Lines beginning with a single hash '#' are treated as comments. See 127 pow-inputs for an example of an input file. 128 129Multiple execution units per function: 130===================================== 131 132Some functions have distinct performance characteristics for different input 133domains and it may be necessary to measure those separately. For example, some 134math functions perform computations at different levels of precision (64-bit vs 135240-bit vs 768-bit) and mixing them does not give a very useful picture of the 136performance of these functions. One could separate inputs for these domains in 137the same file by using the `name' directive that looks something like this: 138 139 ##name: 240bits 140 141All inputs after the ##name: 240bits directive and until the next `name' 142directive (or the end of file) are part of the "240bits" benchmark and 143will be output separately in benchtests/bench.out. See the pow-inputs file 144for an example of what such a partitioned input file would look like. 145 146It is also possible to measure latency and reciprocal throughput of a 147(partial) trace extracted from a real workload. In this case the whole trace 148is iterated over multiple times rather than repeating every input multiple 149times. This can be done via: 150 151 ##name: workload-<name> 152 153where <name> is simply used to distinguish between different traces in the 154same file. To create such a trace, you can simply extract using printf() 155values uses for a specific application, or generate random values in some 156interval. See the expf-inputs file for an example of this workload mechanism. 157 158Benchmark Sets: 159============== 160 161In addition to standard benchmarking of functions, one may also generate 162custom outputs for a set of functions. This is currently used by string 163function benchmarks where the aim is to compare performance between 164implementations at various alignments and for various sizes. 165 166To add a benchset for `foo': 167 168- Add `foo' to the benchset variable. 169- Write your bench-foo.c that prints out the measurements to stdout. 170- On execution, a bench-foo.out is created in $(objpfx) with the contents of 171 stdout. 172 173Reading String Benchmark Results: 174================================ 175 176Some of the string benchmark results are now in JSON to make it easier to read 177in scripts. Use the benchtests/compare_strings.py script to show the results 178in a tabular format, generate graphs and more. Run 179 180 benchtests/scripts/compare_strings.py -h 181 182for usage information. 183